Site Search
Computer Science


Christie Ezeife, Ph.D.Dr. Christie Ezeife
Dr. Christie Ezeife
Dr. Luis RuedaDr. Luis Rueda
Dr. Luis Rueda
Dr. Robert KentDr. Robert Kent
Dr. Robert Kent
Windsor WaterfrontWindsor Waterfront Park
Windsor Waterfront Park
Dr. Ziad Kobti lecturingDr. Ziad Kobti
Dr. Ziad Kobti
Robin Gras, Ph.D.Dr. Robin Gras
Dr. Robin Gras
Arunita Jaekel, Ph.D.Dr. Arunita Jaekel
Dr. Arunita Jaekel
Imran Ahmad, Ph.D.Dr. Imran Ahmad
Dr. Imran Ahmad
Dr. Scott GoodwinDr. Scott Goodwin
Dr. Scott Goodwin
Alioune Ngom, Ph.D.Dr. Alioune Ngom
Dr. Alioune Ngom
Jessica Chen, Ph.D.Dr. Jessica Chen
Dr. Jessica Chen
Xiaobu Yuan, Ph.D.Dr. Xiaobu Yuan
Dr. Xiaobu Yuan
Lambton TowerLambton Tower
Lambton Tower

Trajectory-based Human Action Recognition

Add this event into your calendar using the iCAL format
  • Wed, 05/16/2018 - 1:00pm - 3:00pm

Trajectory-based Human Action Recognition

Doctoral Dissertation by

Pejman Habashi

Date: Wednesday, May 16th, 2018
Time: 1:00 pm – 3:00 pm
Location: Erie Hall, G141

Abstract: Human activity recognition is an important and active research area in computer vision that has numerous challenges. Although trajectories have been used for human action recognition by many researchers, they have not been investigated thoroughly, and their full potential has not been put to the test before this work. 

This thesis formally defines three different trajectory extraction methods, namely interest-point based trajectoriesLucas-Kanade based trajectories, and Farnback optical flow based trajectories. Their discriminant power for human activity recognition task is evaluated and a better trajectory shape descriptor is also proposed, which is a superset of the existing descriptors in the literature. Finally, a method is proposed to augment the trajectories with disparity information. The latter is relatively easy to extract from a pair of stereo images, and it is capable of capturing the 3D structure of the scene. To the best of our knowledge, this work is the first to use the disparity information with trajectories for human activity recognition. The proposed disparity-augmented trajectories have improved the discriminant power of traditional dense trajectories by a significant 3.11%.

Thesis Committee:    
Internal Readers:   Dr. Dan Wu and Dr. Jianguo Lu
External Program Reader: Dr. Kemal Tepe
External Examiner: Dr. Faisal Qureshi
Advisor:               Dr. Boubakeur Boufama
Co-Advisor:         Dr. Imran Ahmad
Chair:                   Dr. Hyuk-Jae (Jay) Rhee

See More: